Design And Research On Fast Intra Coding Unit Decision Algorithm Based On HEVC

In recent years, multimedia technology witnessed a great development. Audio, image and video are in a greater need than before. For video data, the number of data of video on the internet is going up. The demand in terms of video of people moves toward high-definition video. Nowadays, 4K TV is coming out in the market. With the advent of HD and UHD video content, JCT-VC put forward the new video coding standard HEVC(High Efficiency Video Coding) in 2013.Through adopting the flexible coding structure and other techniques, it brings about high efficiency coding when compared with H.264/AVC. In the meanwhile, it could save half of the number of coding bits against H.264/AVC.HEVC adopts the quad-tree coding partitioning structure. The number of HEVC intra prediction increases from 9 modes in H.264/AVC to 35 modes. In addition, four depth levels are in use. At each depth level, 35 prediction modes should go through rough mode decision(RMD) and then selected candidates go to rate distortion optimization(RDO) process. Traverse every depth to find the optimal prediction mode achieves high coding efficiency but imposes a very high computational complexity, which impedes the development of HEVC.In order to solve the problems mentioned above, in recent years, lots of algorithms are proposed, which could reduce the computing complexity of HEVC, especially on aspects of HEVC intra coding. Initially, lots of methods learn from previous schemes proposed in H.264/AVC. Some methods were learnt from the optimization method used in intra mode decision in H.264/AVC. To some extent, it has a good result. Then some researchers began to come up with rough mode decision or add most probable mode so as to optimize the process of intra coding. After that, some proposed in terms of transform unit size decision and residual quad-tree transform. Moreover, more and more researchers focus their attention on coding unit size decision in order to make a greater progress than before in the time savings. However, coding unit spatial correlation is not explored enough. Therefore, it could be improved.To reduce the computation complexity of high efficiency video coding(HEVC), a fast coding unit(CU) size decision algorithm based on Correlation Distance(Corr D) is proposed. Correlation distance(Corr D) is applied to compute the degree of coding unit(CU) spatial correlation. The strong or weak spatial correlation with nearby coding units has influences on current coding unit size distribution. By utilizing the proposed method, some useless coding unit(CU) depth levels could be filtered out.Our experimental results show that the proposed algorithm could make a great progress in terms of encoding efficiency. Compared with HM 12.0, the proposed algorithm achieves the average saving of encoding time 33% with the BD-PSNR less than 0.02 d B loss and negligible BD-RATE increase. Also, the proposed algorithm achieves the average saving of encoding time 12% against the state-of-the-art fast algorithm with the BD-PSNR less than 0.01 d B loss and negligible BD-RATE increase. Therefore, our algorithm makes contributions to the implementation of a real-time HEVC encoder.